A unique feature of prostate cancer is the prevalence of clinically latent lesions that do not pose a health threat, and yet are histologically indistinguishable from clinically important lesions. Identification of risk factors that predict advanced disease carries considerable public health significance. We, and others, have provided evidence that the CAG repeat polymorphism in the androgen receptor (AR) gene is associated with advanced prostate cancer risk in US whites. We propose here a combined epidemiologic/molecular study to evaluate the relationship between AR genotypes and prostate cancer risk, by assembling cases and controls among genetically homogeneous Chinese of Singapore, and by screening for somatic mutations in the AR gene. Cases (approximately 400) will be histologically diagnosed incident cases of prostate cancer, aged 45-74 years, occurring among housing estate residents of Singapore in the 6-year period between Jan. 1, 1993 and Dec. 31, 1998. Controls (an aged-matched random sample of 400 men) will be selected from members of the Singapore Cohort Study of Diet and Cancer, an NCI-funded, ongoing population-based cohort of 60,000 Chinese men and women, who are housing estate residents and aged 45-74 years at baseline. In the molecular component, the prevalence of AR somatic mutations in advanced tumors will be determined and characterized, to check whether the distribution of constitutional AR-CAG repeats of patients with somatic AR mutation-positive tumors differ from those with AR mutation-negative tumors. Although AR mutations were considered in the past to occur rarely in prostate tumors, new data indicate that it might occur quite frequently in advanced disease. The mutations so far analyzed, are often of the "gain-of-function" kind, and might contribute to the progression to hormone resistant disease. The main hypothesis to be tested is that shorter CAG alleles code for more active ARs that in turn increase advanced prostate cancer risk by increasing the rate of prostate somatic mutations among onco- and tumor suppressor genes, including the AR gene itself. Results from this study will enable the critical assessment of the role of the AR in prostate cancer predisposition and progression, and will allow the inclusion of the AR gene in a multigenic model of prostate cancer development.